@phdthesis{repoupi136388,
           month = {August},
            year = {2025},
           title = {PENGEMBANGAN MODEL MACHINE LEARNING UNTUK DETEKSI PERILAKU SISWA},
            note = {https://scholar.google.com/citations?user=EWc6SuoAAAAJ\&hl=id

ID SINTA Dosen Pembimbing:
Rizki Hikmawan: 6122897},
          school = {Universitas Pendidikan Indonesia},
             url = {https://repository.upi.edu/},
        abstract = {Penelitian ini mengembangkan model deteksi perilaku siswa berbasis machine learning menggunakan algoritma YOLOv11 untuk meningkatkan objektivitas pemantauan keterlibatan siswa di kelas. Perilaku yang dikenali mencakup membaca, menulis, dan mengangkat tangan sebagai indikator keterlibatan belajar. Metode yang digunakan adalah kuantitatif dengan pendekatan R\&D dan model CRISP-DM, menggunakan SCB-Dataset5 berisi gambar perilaku siswa dalam tiga kategori tersebut. Model YOLOv11 dilatih untuk mendeteksi perilaku secara real-time. Hasil menunjukkan mAP@0,5 sebesar 0,70 dan F1-score rata-rata 0,67. Perilaku mengangkat tangan memiliki akurasi tertinggi (mAP@0,5 = 0,74), sedangkan membaca terendah (0,68) namun masih layak untuk tahap awal. Model menunjukkan stabilitas dan potensi penerapan di kelas nyata. Penelitian ini berkontribusi pada pengembangan teknologi deteksi perilaku siswa di pendidikan, mendukung pemantauan keterlibatan secara objektif, serta berpotensi menjadi dasar sistem pembelajaran berbasis data untuk membantu guru menyesuaikan strategi pengajaran.
\_\_\_\_\_
This study develops a student behavior detection model based on machine learning using the YOLOv11 algorithm to enhance the objectivity of monitoring student engagement in the classroom. The recognized activities include reading, writing, and hand-raising as indicators of learning engagement. The research adopts a quantitative approach with the R\&D method and the CRISP-DM model, utilizing the SCB-Dataset5 containing images of student behaviors in the three aforementioned categories. The YOLOv11 model was trained to detect these behaviors in real time. The results show a mAP@0.5 of 0.70 and an average F1- score of 0.67. Hand-raising achieved the highest accuracy (mAP@0.5 = 0.74), while reading had the lowest (0.68) but remains acceptable at this early stage. The model demonstrates stability and potential for real classroom implementation. This research contributes to the development of student behavior detection technology in education, supporting objective engagement monitoring, and providing a foundation for data-driven learning systems to help teachers adapt their instructional strategies more effectively.},
          author = {Muhammad Sopian, - and Rizki Hikmawan, -},
        keywords = {Deteksi Perilaku Siswa, YOLOv11, Machine Learning, Keterlibatan Siswa, Pembelajaran Berbasis Data.

Student Behavior Detection, YOLOv11, Machine Learning, Student Engagement, Data-Driven Learning.}
}